feat(web): tokenize input corrections and provide for multi-token predictions 🚂 🔪#15897
Open
jahorton wants to merge 1 commit into
Open
Conversation
User Test ResultsTest specification and instructions User tests are not required Test Artifacts
|
5b1c721 to
d558c86
Compare
6d11477 to
fdd65c0
Compare
d558c86 to
7a6f297
Compare
698be9b to
a1ae1cb
Compare
7a6f297 to
221941b
Compare
a1ae1cb to
dc77cca
Compare
dc77cca to
754d792
Compare
jahorton
commented
May 27, 2026
Comment on lines
+619
to
+623
| const suggestionRange = determineSuggestionRange(transition.base.displayTokenization.tokens, tokenization.tokens, (a, b) => a.spaceId == b.spaceId); | ||
| suggestionRange.transitionId = transition.transitionId; | ||
| const corrector = new TokenizationCorrector(tokenization, suggestionRange.tokensToPredict.length, () => true); | ||
| const predictionPrep = determineTokenizedCorrectionSequence(transition, tokenization, new TokenizationResultMapping([match], corrector)); | ||
|
|
Contributor
Author
There was a problem hiding this comment.
This gets the main correction-loop one step closer to integrating multi-token capabilities. It's just a stop-gap in this PR, though.
f5afb06 to
cc902e9
Compare
754d792 to
f9f9b08
Compare
cc902e9 to
bf547ba
Compare
f9f9b08 to
25f7f16
Compare
bf547ba to
d0e11c9
Compare
25f7f16 to
8fa77c0
Compare
9b972f4 to
fbfd88a
Compare
8fa77c0 to
f30651a
Compare
fbfd88a to
0a85adc
Compare
9db4da8 to
f7c1bfa
Compare
…dictions Build-bot: skip build:web Test-bot: skip
f7c1bfa to
1a7bb31
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
For this PR's "baby step", the common
buildCorrectionSequencemethod is reworked to use the search result type provided byTokenizationCorrector. Work in the previous PR - #16023 - ensures that even custom and legacy model types can provide data adhering to the same search result type despite not being able to leverageTokenizationCorrectoritself.The methods leveraging
buildCorrectionSequence, at this point, still only provide a single token for correction to the method. Baby steps.Build-bot: skip build:web
Test-bot: skip